Validation of Real Estate Ads based on the Identification of Identical Images

摘要

In this paper, the development of an algorithm for recognizing unreliable records of real estate objects based on comparative analysis of images is described. The system of preprocessing of records about real estate objects is presented. The task of image analysis within this system is indicated. And the basic requirements for image validation are noted. Photographs of real estate objects are analyzed with the use of pattern recognition theory approaches. Algorithms for detecting and description of image reference points (SIFT, PCS-SIFT, SURF, GLOH, DAISY, etc.) and perceptual hash algorithms have been investigated to identify forged materials. Conclusions are made about the problems and prospects of applying the proposed approach to the validation of real estate ads.